I've been hunting around for a way to safely detect hung and stale NFS mounts, and can't find anything that works in all cases. I'm using version 3.0.4 and 3.0.5.
Here's what I *can* do: * detect unmounted filesystems, and mount them * detect *stale* filesystems, umount and remount them Both of these can be done via storage: actions, or more manual commands: actions. What I can't do is safely detect the case of a truely hung NFS mount. By "hung," I refer to an NFS mount where the server disappeared, and the clients still have shares mounted. This is distinguished from a simply "stale" mount where the server may have rebooted, but has come back online and can service NFS requests again. What I want to avoid is a fork-bomb type problem where cf-agent continually starts copies of small programs to check for a hung mount (like 'df' or 'mount'). I'm running this on Linux boxes, so I can (and do) check /proc/mounts as a non-blocking way to see if something is mounted. However, this won't tell me if the mount is stale; to do that, I can check the return code of 'df' (it returns 1 on a stale mount). Unfortunately, in the case of *hung* mount, it never returns... Anyone have any clever ideas or suggestions? -- Jesse Becker NHGRI Linux support (Digicon Contractor) _______________________________________________ Help-cfengine mailing list Help-cfengine@cfengine.org https://cfengine.org/mailman/listinfo/help-cfengine
